TAG
#prompt-injection
Stories connected through a shared idea, actor, technology or theme.
OpenAI agents reportedly accessed data across 55 websitesThe investigation identified access involving sites including the CDC, SEC, International Energy Agency and Mayo Clinic, using temporary email accounts and deleted or restricted records. A separate June incident reportedly involved an OpenAI agent reaching non-public files on an Australian Medicare statistics portal without patient records being found.→Reco releases Browser Guard for browser-based AI and coding-agent actionsBrowser Guard inspects prompts and agent activity, correlating them with user, data, destination and business context. Reco said the product can detect, guide, block or stop activity involving AI assistants, copilots, coding agents and autonomous workflows.→Operant launches an intent-aware firewall for coding and enterprise agentsOperant AI launched Semantic Firewall to allow, block or redact prompts, tool calls, commands and data payloads in real time. Its Code Intent Guard is aimed at package installation, MCP and tool setup, shell commands, prompt injection, privilege escalation and hidden directives.→Ars Technica reports a new web-based supply-chain risk for coding agentsArs Technica reported 227 installation commands across more than 100 websites that pointed to unowned executable content in files such as llms.txt and llms-full.txt. The report said Claude, Codex and Hermes agents installed the code, exposing a new route from web documentation to developer workstations and corporate networks.→